A LONGITUDINAL ANALYSIS OF EXTRADYADIC INVOLVEMENT IN DATING RELATIONSHIPS
Michael A. Peterman
A dissertation submitted to the faculty of The University of North Carolina at Chapel Hill in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Department of Psychology (Clinical Psychology).
Chapel Hill 2008
Advisor: Donald H. Baucom, Ph.D. Reader: Deborah J. Jones, Ph.D. Reader: Joseph C. Lowman, Ph.D. Reader: Abigail T. Panter, Ph.D.
Michael A. Peterman: A Longitudinal Analysis of Extradyadic Involvement in Dating Relationships
(Under the direction of Donald H. Baucom, Ph.D.)
Extradyadic involvement (EDI) refers to physical or emotional intimacy that takes place outside an existing romantic relationship. When EDI violates relationship standards of exclusivity, infidelity is said to have occurred. Although EDI and infidelity are both fairly prevalent, their underlying causes have not been well understood historically, and many questions remain about why some individuals become extradyadically involved while others do not. Prior research has uncovered several factors that might contribute to EDI, perhaps the most notable of which have been individual attitudes, relationship quality, and contextual opportunity. General trends suggest that forbidding beliefs, high relationship quality, and low contextual opportunity serve to diminish the likelihood of EDI. However, effects have by no means been unequivocal, in large measure because of the methodological limitations of earlier work. Specifically, the predictors of interest have not been appropriately examined within a longitudinal framework, nor have they typically been integrated into a unified theoretical model. Moreover, opportunity and beliefs have almost uniformly been assessed using measures with questionable psychometric properties, whereas relationship quality has been operationalized differently from one study to the next.
validity within a longitudinal framework. Relationship quality was operationalized in terms of commitment, as defined by the Investment Model, and beliefs and opportunity were
evaluated using newly devised scales exhibiting sound measurement properties. As predicted, lower opportunity, along with more forbidding beliefs about extradyadic participation,
diminished the likelihood of subsequent EDI. Importantly, the effects remained even after controlling for the level of extradyadic engagement observed at time 1. In contrast,
TABLE OF CONTENTS
LIST OF TABLES……….viii
Chapter I. INTRODUCTION………. 1
A. Dating Relationships versus Marriage………... 2
B. Attitudes and EDI……….. 3
C. Relationship Quality and EDI……… 10
D. Opportunity and EDI………..19
E. Aims of the Present Study………..26
II. METHOD………..……… 30
A. Study 1 – Scale Construction………..…………... 30
1. Procedure………..……. 30
2. Participants………. 31
3. Measures……… 32
a. Physical Involvement Scale………... 32
b. Extradyadic Involvement Scale………. 32
c. Extradyadic Beliefs Inventory………... 33
d. Extradyadic Opportunity Scale……….. 34
f. Single Attitudinal Item………... 35
g. Sociosexual Orientation Inventory……… 35
h. Moral Traditionalism Scale………35
i. Extradyadic Involvement Scale……….……… 36
j. Dating History………..……….. 36
k. Physical Attractiveness Item………..……… 36
l. Interpersonal Attractiveness Item………..…… 37
m. NEO Personality Inventory – Extraversion Scale………….. 37
n. Brief Sensation Seeking Scale………... 37
B. Study 2 – Longitudinal Analysis………...………… 38
1. Procedure………... 38
2. Participants………. 38
3. Measures………..………….. 39
a. Investment Model Scale………. 39
b. Background and Relationship Characteristics………... 39
III. RESULTS..………..……….. 40
A. Study 1………..………. 40
1. Factor Analytic Strategy……… 40
2. Factor Analysis of the Extradyadic Beliefs Inventory………... 41
3. Factor Analysis of the Extradyadic Opportunity Scale………..45
4. Factor Analysis of the Extradyadic Inappropriateness Scale………. 47
a. Extradyadic Beliefs Inventory………... 49
b. Extradyadic Opportunity Scale……….. 53
c. Extradyadic Inappropriateness Scale………. 55
7. Correlations Among the Different Sub-Scales……….. 56
B. Study 2………..………. 57
1. Data Analytic Strategy………..………. 57
2. Missing Data………..…… 58
3. Descriptive Statistics………..………… 61
4. Findings from the Logistic Regression………..……… 62
5. Influences of Gender and Semester of Data Collection………. 66
6. Effects of the Extradyadic Inappropriateness Scale………... 69
IV. DISCUSSION………..……….. 72
V. APPENDIX A: Physical Involvement Scale……… 82
VI. APPENDIX B: Extradyadic Involvement Scale………..……. 83
VII. APPENDIX C: Extradyadic Beliefs Inventory………. 84
VIII. APPENDIX D: Extradyadic Opportunity Scale………... 86
IX. APPENDIX E: Extradyadic Inappropriateness Scale………... 87
X. APPENDIX F: Single Attitudinal Item……… 91
XI. APPENDIX G: Sociosexual Orientation Inventory……….. 92
XII. APPENDIX H: Moral Traditionalism Scale………..……... 93
XIII. APPENDIX I: Sexual and Dating History………..…….. 94
XIV. APPENDIX J: Physical Attractiveness Item………..….. 95
XVI. APPENDIX L: Brief Sensation Seeking Scale………..……... 97
XVII. APPENDIX M: Investment Model Scale………...……..……… 99
XVII. APPENDIX N: Background and Relationship Characteristics………... 105
LIST OF TABLES
Table 1 Means and Standard Deviations for the EBI Items……… 41
Table 2 Factor Structure of the Extradyadic Beliefs Inventory……….. 43
Table 3 Means and Standard Deviations for the EOS Items………...45
Table 4 Factor Structure of the Extradyadic Opportunity Scale………. 46
Table 5 Item Means and Standard Deviations for EIS Items……….. 47
Table 6 Factor Structure of the Extradyadic Inappropriateness Scale……… 47
Table 7 Correlational Patterns for the EBI Values Dimension………... 49
Table 8 Correlational Patterns for the EBI Justifications Dimension………. 50
Table 9 Correlational Patterns for the EBI Incentives Dimension……….. 50
Table 10 Correlational Patterns for the EBI Peers Dimension………. 51
Table 11 Correlational Patterns for the EOS Opportunity Dimension………. 53
Table 12 Correlational Patterns for the EIS Inappropriateness Dimension………….. 54
Table 13 Interrelationships Among Predictors of EDI………..…... 56
Table 14 Frequency of EDI by Category………..……… 57
Table 15 Descriptive Statistics of Proposed EDI Determinants………... 60
Table 16 Summary of Logistic Regression Results for EBI Dimensions Predicting EDI………... 61
Table 17 Summary of Logistic Regression Results for Values, Opportunity, Gender, and Initial Status Predicting EDI with Five Multiply Imputed Data Sets………... 61
Table 18 Summary of Logistic Regression Results for Commitment, Opportunity, and Values Predicting EDI……….. 62
Table 20 Summary of Logistic Regression Results for Values,
Opportunity, and Initial Status Predicting EDI……….. 63 Table 21 Summary of Logistic Regression Results for Values,
Opportunity, and Initial Status Predicting EDI with Five
Multiply Imputed Data Sets………... 64 Table 22 Summary of Logistic Regression Results for Values,
Opportunity, Gender, and Initial Status Predicting EDI……… 65 Table 23 Summary of Logistic Regression Results for Values,
Opportunity, Gender, and Initial Status Predicting EDI
with Five Multiply Imputed Data Sets………... 66 Table 24 Summary of Logistic Regression Results for Composite
Index, Opportunity, Gender, and Initial Status Predicting EDI………. 68 Table 25 Summary of Logistic Regression Results for Values,
Opportunity, Gender, and Initial Status Predicting EDI
A large body of research indicates that even while engaged in an exclusive relationship, a number of individuals become romantically involved with someone other than their partner. This type of involvement is commonly labeled in the literature as “extradyadic,” signifying that it occurs beyond the bounds of a current relationship or dyad. In the past, extradyadic involvement (EDI) has been defined variably, sometimes referring to physical intimacy, emotional intimacy, or a combination of the two. Undoubtedly, physical and emotional EDI often occur together, but they are nevertheless distinguishable from one another (Allen, Atkins, Baucom, Snyder, & Gordon, 2004). Generally, physical EDI is considered to include sexual intercourse, along with other perhaps less intimate behaviors, such as oral sex, petting, and kissing. It is, therefore, differentiable from emotional EDI, which is characterized by a combination of affective intimacy and sexual feelings but an absence of physical intimacy itself. An additional concept that is tied closely to both emotional and physical EDI is infidelity. Typically, infidelity is considered to represent a type of emotional and/or physical EDI that also violates relationship standards of exclusivity. Stated differently, infidelity occurs when a partner becomes extradyadically involved in some capacity despite relational proscriptions against such behavior.
Thompson, 1984). Surprisingly, though, even with the level of emphasis given to the physical aspects of involvement, their underlying causes are not well understood. For this reason, physical EDI will be the subject of this investigation. This emphasis in no way reflects the assumption that emotional EDI is of lesser importance, but is instead given because of the many significant questions about physical EDI that remain unanswered.
In the interest of addressing some of these questions, the current study tested a longitudinal model of physical EDI for those involved in dating relationships. The model largely consists of variables that have been evaluated previously, but, in the present context, they are either measured or conceptualized differently than in earlier work. These changes, along with the longitudinal modeling framework itself, are motivated to a considerable extent by the limitations of previous studies. Therefore, in addition to detailing important findings, the following review identifies the specific shortcomings in the existing literature that were redressed in this investigation. Acknowledgement must be given to Allen et al. (2004), who recently provided a comprehensive and useful analysis of the EDI literature. Some of their organizational strategies are applied in the sections that follow.
Dating Relationships versus Marriages
As noted, this study investigated the physical EDI of persons engaged in exclusive dating relationships, but not of those who are married. The exclusion of the latter group stems from both logistical factors, primarily those relating to participant availability, and the desire to limit sources of extraneous variance within the research sample. Indeed, with regard to EDI, and probably some of its related variables, persons in dating and married
(Allen, 2001), and at least to an extent, this difference likely reflects the entailment of less commitment within a dating relationship than a marriage (Forste & Tanfer, 1996). Also, some theoreticians (Brown, 1991; Pittman, 1989; and Reibstein & Martin, 1993) posit the existence of certain developmental stages within marriage, such as the period following the birth of a child, that increase the likelihood of EDI, and this idea has received some empirical support (Allen, 2001). Thus, for each type of relationship, unique factors might influence extrarelational behavior, such that a theoretical model accounting for EDI in one type of relationship might not be equally applicable to the other. Despite these potential differences, the current study was motivated to a significant extent by findings from the marital literature, and hence these are reviewed below along with other relevant work. Importantly, though, marital study interpretation should not overlook the possibility that EDI may differ as a function of relationship type. That being noted, a divergent point of view acknowledges the possible similarity between dating and marital relationships with respect to extrarelational behavior. Central to this perspective is the notion that during dating and courtship,
individuals develop relationship scripts that carry over into marriage (Hansen, 1987; Wiederman & Hurd, 1999; Roscoe, Cavanaugh, & Kennedy, 1988), and although this position will not be tested here, it is important to acknowledge the potential similarities of EDI across relationship types, along with its possible differences.
Attitudes and EDI
part, a reliable level of concordance between EDI attitudes and behaviors has been observed empirically, but due primarily to issues of methodology, prior studies have done little to establish a causal link between attitudes and actual involvement. Here, relevant findings are presented, along with a review of the methodological limitations of earlier work.
A common approach to evaluating the relationship between EDI attitudes and behavior has involved the extraction of pertinent data from large national surveys. For example, in an effort to evaluate both the incidence and correlates of EMI, Wiederman (1997) analyzed data from a general social survey that, among many other issues, asked participants to concurrently report on their history of, and attitudes toward, extramarital involvement. Involvement history was coded dichotomously on the basis of whether participants reported “having sex” with someone other than their spouses while married. Attitudes toward the permissibility of EMI were solicited through the single item, “What is your opinion about a married person having sexual relations with someone other than the marriage partner—is it always wrong, almost always wrong, wrong only sometimes, or not wrong at all?”
while attitudes were correlated with actual extradyadic involvement, the two were by no means entirely in concert with one another.
Using an approach similar to that of Wiederman (1997), other investigators have also found evidence supporting a relationship between EDI attitudes and behavior. Choi, Catania, and Dolcini (1994) used national survey data assessing the incidence of extramarital sex within the last 12 months, along with beliefs about monogamy. Respondents rated their level of agreement with the statements, “Sometimes it is okay for married people to have sex outside their marriage,” and “Having sex with someone other than your husband/wife is always wrong.” Consistent with Wiederman’s (1997) findings, stronger beliefs in the
importance of monogamy were associated with lower levels of EMI within the preceding 12 months. The effect was independent of respondent ethnicity. Using national survey data and a single item to assess EDI attitudes, Treas and Giesen (2000) observed a similar relationship between EDI beliefs and behavior, one that remained even after the effects of other important variables, such as opportunity and relationship satisfaction, were statistically controlled. Together then, the results derived from survey data are indicative of a fairly reliable and perhaps unique association between EDI attitudes and behavior. However, because of methodological considerations, the findings from each of these studies must be interpreted with qualification.
First, in each study, attitudes and behavior were assessed concurrently, thus precluding a conclusive judgment of whether beliefs actually determine the level of
because the items ask about the behavior of a hypothetical other and not the respondents themselves, their validity is questionable, given the primary interest in participants’ attitudes about their own behavior. Together, these factors militate against a clear understanding of whether beliefs about EDI influence its actual occurrence.
Different methodology has been used elsewhere. Rather than using only a single item to assess attitudes toward EDI, Buunk and Baker (1995) employed an eight-item scale on which respondents indicated their level of agreement with four statements that approved of extramarital relations and four that disapproved. The investigators also departed from others by selecting the willingness to engage in EMI, should the opportunity present itself, instead of actual engagement as their dependent variable. Even with these differences, the findings were largely consistent with those of other studies. Specifically, attitudes were significantly predictive of extradyadic sexual willingness, accounting for approximately 8% of the variance among participants in one sample and 32% among participants in a second sample assessed 15 years after the first. As predicted, respondents with tolerant attitudes were more willing than those opposed to EDI to become involved extradyadically. However, the findings do not indicate how closely EDI attitudes relate to actual behavior, which is probably not equivalent to the mere willingness to become extradyadically involved.
one’s career. Respondents were asked to indicate on a four-point scale the degree to which each factor would justify their own extramarital involvement. It is noteworthy that relative to a single item, this measure likely has more discriminative power. Presumably, a respondent not justifying EMI for any of the reasons given would be less permissive than one justifying such behavior on some grounds but not others. If so, the measure might be more sensitive to attitudinal differences between respondents.
To assess EMI itself, the authors inquired about a continuum of behaviors, including kissing and caressing, rather than sexual intercourse exclusively. As predicted, both men and women who justified EMI along the sexual dimension were more likely to report a history of sexual EMI. Moreover, for men but not women, justification along the love dimension was significantly correlated with sexual EMI. Therefore, with both an attitudinal measure of presumably greater reliability and the assessment of a broader range of extradyadic behaviors, more accepting attitudes toward EMI were significantly related to actual involvement. While this further supports the influence of attitudes on behavior, the largest effects in the study were only moderate in size, indicating that much of the variance in sexual EMI was unrelated to EMI justifications. Also, as in other studies, attitudes and extramarital involvement were measured concurrently, precluding any meaningful assessment of the potential causal relationship of attitudes to involvement.
In addition to those who are married or cohabiting, attitudes also predict EDI for persons in dating relationships, although this population has admittedly garnered only scant attention in earlier work. Among individuals in “exclusive” relationships, characterized by the expectation of sexual and dating infidelity, those with unrestricted sociosexual
(Seal, Agostinelli, & Hannett, 1994). Generally, sociosexuality refers to a person’s willingness to engage in uncommitted sexual behavior (Simpson & Gangestad, 1991). Compared to restricted persons, those unrestricted are more likely to have sex earlier in a relationship, to report a history of multiple sex partners, and to engage in sexual relationships not characterized by love, commitment, involvement, and dependency (Simpson &
Using both a self-report measure and a behavioral experiment, Seal et al. (1994) examined the willingness of individuals to engage in extradyadic behaviors with attractive opposite-sex partners. For the behavioral component, respondents were given the option of pursuing a date with someone who appeared briefly in a mock video. Both in terms of their own self-report and their actual behavior, those with unrestricted sociosexual orientations, relative to those with less restricted orientations, were more willing to engage in extradyadic behavior. Additional evidence corroborates this finding (Wiederman & Hurd, 1999). College students who dissociated sex, love, and marriage, along with those who were more sexually sensation seeking, reported higher rates of extradyadic activity. Together, these findings suggest that for those in “exclusive” dating relationships, permissive sexual attitudes promote EDI, but the results also raise the question of whether permissive sexual attitudes in general equate to a more liberal perspective regarding extradyadic behavior.
extradyadic involvement relative to those more tolerant of EDI. Thus, while evidence for those in dating relationships is limited, there exists at least a preliminary indication that irrespective of relationship type, attitudes about EDI influence the level of actual involvement.
Together, available findings support a number of broad conclusions each critical to an understanding of the relationship between EDI behavior and attitudes. To be certain, attitudes about EDI and involvement are fairly reliably correlated, and individuals who are more accepting of EDI generally report higher levels of involvement compared to those with less permissive beliefs. However, EDI behaviors and attitudes are by no means entirely
concordant. Typical effects are only modest in size, suggesting that individuals do not always behave in congruence with their beliefs. Also, because the two have always been examined concurrently, there exists no clear indication that EDI beliefs actually engender different levels of extradyadic behavior. A longitudinal investigation would address the latter point, clarifying whether beliefs expressed at one point influence subsequent levels of extradyadic activity. As for the seemingly modest correlation between attitudes and involvement, it might accurately reflect the true strength of the relationship in question. However, effects may have been attenuated by methodological limitations, most notably the low reliability and
more direct measure might also help to clarify the proposed relationship of EDI attitudes to involvement.
Related to issues of measurement, it might also be necessary to more broadly conceptualize EDI attitudes themselves. In past studies, attitudes have often been defined narrowly, usually reflecting a simple judgment about the acceptability of extradyadic behavior. The tendency ignores potential variability both in the form and nature of EDI beliefs. For example, along with general opposition to extradyadic activity, an individual may hold the more fundamental belief that anyone engaging in EDI is necessarily
unscrupulous and immoral. Furthermore, some individuals may see EDI as the potential cause of aversive outcomes, such as damaged self-concept, alienation from friends, and emotional distress. When considered together, such a network of cognitions might better encapsulate an individual’s underlying beliefs. If so, it should predict actual behavior more robustly than a simple attitudinal measure.
Relationship Quality and EDI
In addition to individual factors, such as EDI attitudes, relationship variables have been evaluated as possible antecedents to extradyadic involvement. In large part, this line of research has been organized around the position that EDI is more likely to occur when
individuals feel distressed in, or otherwise displeased with, a primary relationship. Generally, extant findings support this broad conclusion. Important results are reviewed here, both in terms of their substantive relevance and implications for future research.
attitudes, many investigators have relied on data from surveys as a means of assessing relationship quality, which usually is the subject of only one or a few items, and concurrent levels of extradyadic involvement. Using this approach with a sample of individuals who were either married or cohabiting, Treas and Giesen (2000) defined overall relationship satisfaction in terms of the emotional fulfillment and physical pleasure derived from the union. Importantly, although respondents differed considerably in the duration of their relationships, satisfaction data were only available for those establishing their unions within the prior 12 months. For this subset of participants, relationship dissatisfaction was positively associated with recent extradyadic sexual intercourse, and this effect remained significant even after the effects of attitudinal and opportunity variables were statistically controlled.
Utilizing a similar approach, Atkins, Baucom, and Jacobson (2001) asked participants to categorize their primary relationships as “very happy, pretty happy, or not too happy.” Consistent with their hypothesis, level of satisfaction was significantly related to the incidence of extramarital sex (EMS). Indeed, as compared to those who described their relationships as “very happy”, those with “pretty happy” and “not too happy” unions were two and four times as likely, respectively, to report EMS. Glass and Wright (1977) also measured relationship satisfaction through a single item asking respondents to rate the overall quality of their marriage on a scale ranging from “very happy” to “very unhappy.” Marital satisfaction was significantly associated with extramarital sexual involvement, with less satisfied partners reporting more extramarital behavior. In a subsequent study, the investigators replicated this earlier finding but further evaluated the association of
Notably, the relationship between the two was more pronounced for women than men, an indication that the influence of relationship quality on EDI may differ by gender.
This interaction has been supported elsewhere, albeit on the basis of a slightly different research design (Buss & Shackelford, 1997). Rather than reporting their actual extramarital involvement, participants estimated the likelihood of their eventual engagement in a variety of physical extradyadic acts. As in Glass and Wright (1985), marital satisfaction was negatively correlated with the estimated likelihood of future EDI, and the strength of this correlation was greater for women than men. Allen et al. (2000) also found a larger effect of relationship satisfaction on EDI for women relative to men. Thus, not only do available findings support the general link between relationship quality and EDI, they also provide converging evidence of a stronger association for women than men. Overall, though, the more general association has received wider support.
EDI. Indeed, those reporting a history of extradyadic sex are typically less satisfied with coital frequency and/or quality in their primary relationships (Traeen & Stigum, 1998; Bell, Turner and Rosen 1975; Spanier & Margolis, 1983; Thompson, 1983). Moreover,
extradyadic sex is less common among those who are more emotionally satisfied with their primary sexual relationships and among women who derive more physical pleasure from those relationships (Waite & Joyner, 2001).
Disparities of equity, which certainly might influence relationship quality, have also been evaluated as possible contributors to EDI. Briefly, an equitable relationship is often defined as one that is equally beneficial to both partners. According to equity theory,
inequitable relationships almost invariably cause both the under- and over-benefited partners to feel distressed, and their distress encourages them to restore equity within the relationship. On these theoretical grounds, Walster, Traupmann, and Walster (1978) reasoned that in an attempt to establish equity, under-benefited partners might participate in extramarital relations. To evaluate this prediction, married and cohabitating individuals were asked to concurrently report on their relationship equity and level of extradyadic involvement. As expected, under-benefited partners, in comparison to both the over-benefited and those in equitable relationships, admitted to more extradyadic encounters. Furthermore, the under-benefited partners became extradyadically involved earlier during the course of their primary relationships. Additional evidence tacitly suggests that the appeal of a more equitable
relationship promotes extradyadic activity, as those engaging in EMI report greater equity in their extradyadic relationships as compared to their marriages (Glass, 2003).
men, and interestingly, both under- and over-benefited women have been more inclined to engage in EDI, a finding that diverges from the prevailing assumption that extrarelational activity is more common among those under-benefited in their primary relationships (Prins, Buunk, & VanYperen, 1993). These effects were independent of both relationship and sexual satisfaction, in addition to individual attitudes about EDI. It does, however, bear mentioning that relative to relationship factors, attitudes were more strongly associated with both the desire for, and the engagement in, extrarelational activity. Thus, the importance of inequity relative to other factors, along with its effects across gender, are somewhat unclear. Further obfuscating these issues, Prins et al. (1993) operationalized EMI as the number of
extramarital partners, a measurement strategy that probably lacks precision. For these reasons, the influence of equity on extradyadic involvement requires additional clarification.
Related to equity is the distribution of relationship power, an imbalance of which may promote EDI. On the basis of who usually prevails in disagreements, Edwards and Booth (1976) classified relationships as male- or female-dominated. Although a disparity in power had a negligible effect for men, women dominant in their relationships were more likely to report EMI. Using employment status as an indicator of power, Atkins et al. (2001) found similar results, although without any moderation by gender. For both men and women, a discrepancy in spouses’ employment status predicted greater extramarital involvement, although, as noted by the authors, this finding could also reflect an increase in EMI
opportunity following status ascension. Variables of opportunity will be discussed in greater detail in a later section.
Typically, relational factors have been examined individually for their effects on EDI. In contrast to this trend, however, the impact of relationship variables has also been evaluated within the context of a broader and more unifying theoretical framework (Drigotas, Satstrom, & Gentilia, 1999). Based on the assertion that infidelity is principally a function of
relationship commitment, the authors utilized an investment model of commitment, originally proposed by Rusbult (1980, 1983), to predict EDI longitudinally. Within this model, commitment, defined as the intent to persist in a relationship, has three bases of dependence: (a) satisfaction, which relates to the outcomes derived from a relationship; (b) investment, referring to those things that would be lost were the relationship to terminate; and (c) alternative quality, as represented by the most attractive alternative to the
More so than its underlying determinants, commitment is thought to influence relationship behavior directly, affecting decisions to persist and fostering a long-term orientation to the involvement. Furthermore, commitment seems to elicit pro-relationship transformation (Rusbult, Martz, & Agnew, 1998), a process in which one’s own self-interest is subjugated to the needs of the relationship (Holmes, 1981; Kelley & Thibaut, 1978). Indeed, commitment is associated with a number of actions beneficial to a relationship, including tendencies to accommodate a partner who has behaved negatively (Rusbult,
committed individuals would probably eschew involvement with another in an effort to protect the primary relationship, possibly to the detriment of their own immediate self-interest. Under this rationale, infidelity would stem largely from an erosion in commitment (Drigotas et al., 1999).
To test the claim, undergraduates in dating relationships provided data on two occasions, during the first assessment reporting overall commitment in addition to levels of satisfaction, investment, and alternative quality specifically. Data on EDI were collected at a second point, roughly two months after the first; at the second time period, respondents indicated their level of extradyadic involvement since the first point of assessment. Rather than inquiring about EDI in general, the authors restricted their measure, asking only about involvement with the person other than the primary partner to whom the respondent was most attracted.
Regression analyses were used to determine if the commitment variables, as measured at time 1, were predictive of the level of EDI reported at time 2. Consistent with the hypothesis, time 1 commitment and its constituent dimensions of satisfaction, investment, and alternative quality were each significantly predictive of a composite EDI variable.
However, for physical intimacy specifically, only commitment and satisfaction were
This conclusion, however, should be qualified for at least three reasons. First, in assessing EDI, the investigators only considered involvement with the person other than the primary partner to whom the respondent was most attracted. This assessment strategy is questionable given the potential for extradyadic involvement with other individuals as well. Additionally, EDI was not measured during the initial assessment. Therefore, although less committed individuals at time 1 were more inclined to report EDI at time 2, initial levels of EDI were never established. Without any knowledge of the extent to which respondents became more or less extradyadically involved over the course of the study, it is impossible to determine if commitment was significantly predictive of changes in EDI. Lastly, the effects of other potentially influential variables, such as EDI attitudes, were not included in the study. Thus, neither the unique impact of commitment nor its interactions with other factors could be evaluated for their effects on extradyadic involvement.
Despite the limitations, though, the investment model might be instrumental to a better understanding of EDI development. Within the model, a number of earlier findings can be usefully integrated, an important consideration given the lack of theoretical cohesion that is evident in many prior studies. For example, as a specific dyadic problem, sexual
dissatisfaction could erode relationship satisfaction in general and possibly enhance the perceived appeal of alternative partners offering a more sexually fulfilling relationship. As a result, commitment itself might be lessened, with extradyadic behavior then becoming more likely. To further the argument, a power disparity might also undermine relationship
satisfaction, with the less powerful partner perhaps feeling undervalued and the more powerful partner feeling underbenefited in the union. Under either of these scenarios,
and increasing the likelihood of EDI. In this sense, commitment would mediate the effects of other relationship factors on EDI. Interestingly, Atkins (2003) evaluated the utility of both global and specific indicators of relationship satisfaction in the prediction of extradyadic involvement. He concluded that after controlling for the effects of more specific factors, global measures of satisfaction were not uniquely tied to EDI. These findings suggest that global indicators might serve as mere proxy variables for more specific relationship
dynamics, which may have greater precision as predictors of extradyadic involvement (Allen et al., 2005). Nevertheless, because of its theoretical cohesion, the investment model is heuristically useful, and its validity should be examined further.
Overall, available findings converge to suggest that diminished relationship quality promotes extradyadic involvement, and this is perhaps more so the case for women than men. Global measures of distress are typically associated with EDI, as are more specific indicators such as sexual dissatisfaction and relationship inequity. However, EDI and relationship quality typically have been evaluated cross-sectionally, which is problematic for at least two reasons. First, the disclosure or discovery of a partner’s EDI could severely diminish
relationship quality, making it impossible to determine if relational distress preceded EDI or vice versa. Likewise, there might be mutually reciprocal influences between EDI and
relationship quality. In one study, divorce proneness predicted extramarital sex, the
relationship quality and later extradyadic behavior. For these reasons, the abundance of cross-sectional designs precludes inferences of causality, an important consideration given the objective of explaining EDI’s development. Existing studies are also marked by an apparent lack of theoretical cohesion, although this might be partially rectified with the investment model. Its utility, however, requires further evaluation, particularly since it has not yet been evaluated within the context of other predictors. Additional explanatory variables might help to explain why, for example, those less committed sometimes eschew EDI, instead terminating a current relationship before pursuing an alternative partner. On the whole then, a longitudinal research design with multiple predictors would effectively remedy important shortcomings of earlier work.
Opportunity and EDI
While the investment model largely comprises relational factors, it is also
contextually dependent, particularly in the sense that alternative quality can be influenced meaningfully by factors external to the relationship. Its possible association to alternative quality suggests that EDI is in part a function of situational opportunity, a position supported by a large body of research. More specifically, a host of studies support the notion that as the opportunity to do so increases, individuals are more likely to engage in extradyadic behavior. However, despite a fairly robust association with EDI, opportunity has most often been measured indirectly through indicating factors with questionable validity. This, along with the fact that it has usually been measured concurrently with EDI, precludes a clear
understanding of whether opportunity promotes extradyadic behavior.
Rather than measuring it directly, the authors identified factors that would reasonably promote opportunity. These included the individual ability to attract sexual partners, a job requiring personal contact with potential sex partners, an urban residence, and a social
network approving of EDI. The number of sexual partners between age 18 and the first union served as a proxy for the ability to attract partners. To assess the degree of personal contact with potential partners in the workplace, respondents were asked to report on how much they touched, talked to, or spent time alone with others on the job. Urban residence was coded dichotomously on the basis of whether participants lived in a large- or medium-sized central city, and finally, approving social networks were assessed in terms of religious attendance and the extent to which respondents enjoyed spending time with a spouse’s family and friends. The unique effect on EDI of each of these factors was evaluated after controlling for attitudinal and relationship variables.
Consistent with prediction, the number of previous sexual partners was positively related to EDI, as was a low degree of overlap in spouses’ social networks. Furthermore, for the prior 12 months only, greater religious attendance and lower personal contact within the workplace were predictive of less EDI. On the surface then, it seems that opportunity is significantly related to extradyadic involvement, but it must be noted that, in the above study, opportunity was evaluated indirectly. Thus, the overarching conclusion of a significant relationship between opportunity and EDI is a tentative one at best. Indeed, on a number of grounds, the validity of the opportunity measures can be reasonably questioned. For example, a respondent with an extensive sexual history might not continue to attract potential sex partners after marriage. Perhaps acculturation to family life, among other factors,
surmise that social settings other than the workplace provide access to potential extradyadic partners. Therefore, in the absence of a more pointed measure of opportunity, its relationship to EDI is unclear, a problem that is further amplified by the concurrent measurement of the variables, which precludes any inferences relating to causality.
Despite its limitations, the methodology of Treas and Giesen (2000) is common among studies investigating the influence of opportunity on extradyadic involvement. Indeed, urban residence has often been used as a proxy for opportunity, although a consensus has not yet emerged as to whether larger communities actually promote EDI. Similar to the results of Treas and Giesen (2000), some findings support such a relationship. Traeen and Stigum (1998) measured population density on a continuum ranging from sparse towns with fewer that 200 inhabitants to large cities with more than 100,000 inhabitants. Respondents from more densely populated areas showed a greater inclination to engage in extradyadic sex, a finding corroborated elsewhere (Kontula & Haavio-Mannila, 1995).
However, support for this association is not universal. Choi et al. (1994) found no difference in the incidence of EDI over the prior 12 months when comparing urban
respondents to those from a national sample, although the legitimacy of this comparison is dubious given that many from the national sample might have resided in urban areas. Moreover, in a reevaluation of the data, Allen et al. (2004) concluded that for men, urban residence may actually have been predictive of greater EDI. Therefore, the initial findings must be qualified in terms of these key points. However, the absence of a significant
12 largest metropolitan areas. For both men and women, community size was unrelated to both the incidence over the prior 12 months and the lifetime prevalence of EDI. Wilson (1995) also found no relation between urbanism and the incidence of EDI over the prior 12 months, concluding that an urban residence may encourage permissive sexual attitudes without promoting actual behavior.
Like urban residence, factors of employment have often served as proxy variables for opportunity. A large percentage of extramarital involvements develop with coworkers
although as others note (Allen et al., in press), this may stem from oversampling on the basis of urban residence and minority status, both of which correlate with higher rates of EDI.
Similar to a higher income, extensive travel might allow for surreptitious behaviors facilitative of EDI. While not evaluated extensively, there does appear to be some association between travel and extradyadic behavior. A history of EDI is related to the number of
traveling days in a year (Traeen & Stigum, 1998), although the strength of this finding is unclear, particularly since the effects of other variables, such as EDI beliefs and overall relationship quality, were not controlled. Perhaps, for unknown reasons, those with
permissive sexual attitudes gravitate towards jobs requiring travel, or perhaps the selection of such employment stems from problematic aspects of the primary relationship. Despite these possibilities, others have observed a similar association between travel and EDI. Specifically, Wellings, Field, Johnson, and Wadsworth (1994) found that multiple sexual partners were more common among those whose jobs required overnight travel.
be pursued with minimal threat of discovery by the primary partner. This supports the more general position that opportunity promotes extradyadic behavior.
However, in no reported case has opportunity been measured directly in terms of its theoretical definition, a limitation that could account for inconsistencies in available findings as well as the generally unimpressive size of effects. On the surface, this measurement strategy seems puzzling, although Atkins et al. (2001) astutely note one possible reason for its use. With a cross-sectional design, those reporting extradyadic involvement would necessarily endorse higher levels of opportunity. Thus, without the use of proxy variables that might estimate the level of opportunity prior to involvement, the two constructs are inextricably connected, such that the direct assessment of opportunity would likely overestimate its causal affect on EDI. The combination of a longitudinal design with appropriate statistical methodology would remediate this problem, though, as EDI could be predicted from an earlier, direct measure of opportunity, after controlling for the effects of an earlier measure of EDI. This approach might clarify inconsistencies in the current literature and better estimate the true impact of opportunity on subsequent extradyadic behavior.
some to consider individual factors, such as the ability to attract extradyadic partners, which might lead to greater opportunity for extradyadic behavior (Treas & Giesen, 2000). The current study adopted a similar approach to evaluate further individual-level variables that might promote or discourage level of opportunity.
Aims of the Present Study
As stated earlier, the current investigation was motivated not only by the results, but also the limitations, of prior research. In terms of results, a number of general conclusions have been supported. First, available findings suggest that extradyadic involvement is more common among those with permissive attitudes. Additionally, poor relationship quality seems to predict a higher incidence of EDI, as does increased opportunity. Of importance, associations between EDI and its predictors seem fairly robust, as each has been observed consistently and over a range of methodology. However, almost invariably, data have been gathered concurrently, and as a result, the causal influence of each predictor on EDI is not well understood. Therefore, a longitudinal study would contribute substantially to the existing literature.
Among the limitations of earlier studies, perhaps most important is the general
absence of a unifying model within which to conceptualize the onset of extradyadic behavior. Of notable exception is the work of Drigotas et al. (1999), in which the emergence of EDI was organized usefully around the investment model of relationship commitment. However, the investment model seems incomplete in its explanation of extrarelational involvement, at least in part because it fails to account for EDI attitudes. Moreover, the model seems to equate opportunity and alternative quality, even though the two are quite possibly dissimilar. To fully account for the occurrence of EDI, a more expansive model is likely necessary. To redress various issues that have been detailed, the current study tested a
longitudinal model of dating EDI that included situational opportunity, individual beliefs, and relationship commitment as predictor variables. As formulated, the theoretical model stipulates that predictors operate jointly to influence extradyadic participation. It is assumed that EDI can only occur when opportunities exist for accessing available and willing
alternative partners. As detailed earlier, opportunity itself is likely increased through engagement in certain interpersonal scenarios, such as flirting with an attractive person. Indeed, the engagement in such scenarios might initiate a trajectory toward EDI, in which more casual interactions gradually intensify, thereby creating attraction for an alternative partner along with opportunities for physical engagement. When opportunities arise, the decision of whether to pursue extradyadic relations depends upon the remaining variables in the model, namely the individual’s beliefs about EDI along with the level of commitment to the primary relationship. If EDI beliefs are forbidding and/or commitment is high,
both. However, even under these conditions EDI may still occur, particularly if the available opportunity is highly appealing. When EDI beliefs are permissive, opportunities will be pursued more readily, given that extradyadic engagement produces no discrepancy between attitudes and behavior. Similarly, less committed individuals, perhaps unconcerned with maintaining a primary relationship, will more frequently involve themselves extradyadically in response to increased opportunity. However, the effects on EDI of both commitment and beliefs are seen as interactive, such that an individual with permissive attitudes might eschew EDI if commitment is high, just as a person with highly forbidding beliefs might disavow EDI even when commitment is low.
To further address the limitations of prior studies, a number of measurement issues received consideration in the present investigation. Rather than relying on the ordinal or dichotomous measures of earlier research, we attempted to derive a continuous scale of extradyadic involvement. EDI attitudes were broadly defined with an original scale that more fully represented underlying beliefs about extradyadic participation. Furthermore,
opportunity was also measured on a new scale, one arguably yielding a more valid assessment than the proxy variables used elsewhere.
Summary of Hypotheses
As proposed, the current study directly tested the following hypotheses:
(1) Negative beliefs about EDI will predict lower levels of extradyadic involvement. Thus, as compared to individuals with permissive EDI beliefs, those expressing more negative
(3) Beliefs and commitment will interact to influence EDI. Specifically, as beliefs become more forbidding, the impact of commitment on EDI will diminish. Likewise, as commitment increases, the influence of beliefs on EDI will decrease.
In the interest of clarity, the following sections make a distinction between methods of scale construction and those associated with the full longitudinal analysis. However, scale development actually occurred as a specific component of the broader longitudinal
investigation. Following a pilot test with a separate group, undergraduates in the full study completed measures on four occasions, with the first yielding all necessary data for scale development. Therefore, although the psychometric and longitudinal aspects of the study are described separately, each relied upon a common group of research participants.
Study 1 – Scale Construction Procedure
Study 1 pilot tested and psychometrically evaluated three newly devised measures: the Extradyadic Beliefs Inventory (EBI), Extradyadic Opportunity Scale (EOS), and Extradyadic Inappropriateness Scale (EIS). The EBI and EOS were created to assess EDI beliefs and opportunity level, respectively, whereas the EIS was designed to measure
intimacy level of various dyadic physical behaviors. The ratings facilitated construction of a continuous measure of physical extradyadic involvement. Data collection occurred via the web through the Zoomerang Pro software package.
Overall, 60 participants completed the pilot test and an additional 318 enrolled in the full study, which involved both a psychometric analysis of the newly created scales and the longitudinal investigation of EDI. Data collection occurred over two academic semesters at UNC-Chapel Hill. During the first, which included pilot testing, all participants were engaged in an exclusive dating relationship in which any extradyadic physical engagement was forbidden. For the second semester, inclusion criteria were relaxed somewhat to allow participation of individuals engaged in non-exclusive dating relationships. The modification stemmed from the relatively sparse occurrence of extradyadic engagement during the first semester. Across both periods of data collection, married persons were excluded.
Furthermore, to prevent nesting in the data, students could not participate if their romantic partner was doing so. Recruitment approximately balanced the number of participants across gender. No additional selection criteria were imposed.
Physical Involvement Scale (PIS; see Appendix A). On the PIS, participants used a 100-point scale to rate the level of physical involvement associated with the following behaviors: (a) passionate kissing; (b) sexual hugging and caressing; (c) heavy petting; (d) oral sex or other similar sexual contact; and (e) sexual intercourse. For example, an individual might have assigned a value of 90 for the physical involvement of sexual
intercourse, and a value of 45 for heavy petting. As described, sexual intercourse would have involved twice the level of physical engagement as heavy petting. For each behavior, the average rating across participants contributed to the development of a continuous measure of extradyadic involvement, in which behaviors were weighted by their level of physical
intimacy. Related procedures are reviewed more fully in the method section of Study 2. Extradyadic Involvement Scale (EDIS; see Appendix B). A subset of items from the Extradyadic Experiences Questionnaire (EEQ; Allen and Baucom, 1999) measured
extradyadic involvement. The items themselves asked about the frequency and occurrence of a full range of physical dyadic behaviors, including passionate kissing, sexual hugging and caressing, heavy petting, oral sex or other similar contact, and sexual intercourse. The resulting data, along with average intimacy ratings from the PIS, permitted construction of a continuous measure of EDI, with three approaches receiving consideration. (1) The first summed the dichotomous codes for each behavior, resulting in a continuous score ranging from 0 to 5. (2) A second approach weighted behaviors by their PIS intimacy levels. Clearly, some behavioral categories subsumed others. For example, an individual engaging in
to determine an EDIS score. (3) A final approach weighted intimacy both in terms of frequency and PIS score.
Newly Constructed Measures. The Extradyadic Beliefs Inventory, Extradyadic Opportunity Scale, and Extradyadic Inappropriateness Scale were developed to evaluate predictors included in the theoretical model. Each is described below.
Extradyadic Beliefs Inventory (EBI; see Appendix C). As compared to prior work, the present study defined EDI beliefs rather broadly. Specifically, here beliefs relate not only to the perceived rightness or wrongness of EDI, but also to its anticipated consequences. As formulated, permissive attitudes and a low expectation of negative consequences would lead to greater EDI tolerance, whereas forbidding attitudes and a high expectation of aversive consequences would characterize a more oppositional belief structure.
The EBI item pool included 37 statements measuring EDI beliefs, and for each, respondents rated their agreement on a five-point likert-type scale (1 = “strongly disagree”; 5 = “strongly agree”). Because of the emphasis on predicting physical EDI, statements
Convergent and Discriminant Validity. Correlations between the EBI and other selected measures examined convergent and discriminant validity. Measures included a modified version of the single item assessing EMI attitudes, along with indexes of sociosexual orientation, moral traditionalism, and EDI history.
Extradyadic Opportunity Scale (EOS; see Appendix D). Opportunity refers to the availability and willingness of alternative partners, along with factors promoting
secretiveness (Allen et al., 2005). The EOS item pool comprised 15 statements assessing opportunity, each rated on a likert-type scale (1 = “strongly disagree”; 5 = “strongly agree”). Consistent with the conceptual definition, high opportunity signified the availability and willingness of alternative partners, along with situational factors that might prevent discovery of EDI by the primary partner.
Convergent and Discriminant Validity. Along with a measure of extraversion, items relating to dating history, sensation seeking, physical attractiveness, and interpersonal attractiveness examined the convergent and discriminant validity of the EOS.
Extradyadic Inappropriateness Scale (EIS; see Appendix E). For the most part, opportunity has been defined contextually with little regard for individual factors that might promote or discourage it. The current model proposes that some individuals limit access to alternative partners through their avoidance of opportunity-enhancing interpersonal
scenarios. Consistent with the idea, the EIS asked respondents to rate the appropriateness of their engagement in nine hypothetical scenarios (5-point likert-type scale; 1 = “not at all inappropriate”, 5 = “very inappropriate”). Rather than depicting actual extradyadic
Convergent and Discriminant Validity. The same group of measures applied to the EOS also evaluated EIS convergent and discriminant validity.
Single Attitudinal Item (Davis & Smith, 1994; see Appendix F). In much available survey research, EMI attitudes have been assessed through the single item, “What is your opinion about a married person having sexual relations with someone other than the marriage partner—is it always wrong, almost always wrong, wrong only sometimes, or not wrong at all?” Scores range from 0 to 3, with higher values denoting less EDI opposition. As noted, the item is at least moderately related to EMI history, with more permissive attitudes predicting greater extradyadic involvement on the part of the respondent. The EBI differed considerably from the single item, not only because it was self-referenced, but also because it ostensibly measured a broader range of attitudes and beliefs about extradyadic activity. Nevertheless, an overall score on the EBI was expected to correlate significantly with a single item modified to inquire about a dating relationship rather than marriage.
Sociosexual Orientation Inventory (Simpson & Gangestad, 1991; see Appendix G). The sociosexual orientation inventory assesses the willingness to engage in uncommitted sexual relations. Two of its items evaluated convergent and discriminant validity for the newly created scales. The first asked about the frequency of sexual fantasy, whereas the second inquired about agreement level with the statement “Sex without love is okay”.
Because more permissive sexual attitudes in general would likely predict greater tolerance of extradyadic behavior, an unrestricted sociosexual orientation was expected to correlate significantly with more tolerant EDI beliefs.
eight items each scored on a five-point likert-type scale (1 = strongly agree; 5 = strongly disagree). Content areas include sexual freedom, cohabitation, divorce, and traditional family ties (Robinson, Shaver, & Wrightsman, 1999). Internal consistency, convergent validity, and discriminant validity have been established elsewhere (Conover & Feldman, 1985). Morally traditional participants were expected to disapprove of EDI; thus, a positive correlation between moral traditionalism and the EBI was predicted.
Extradyadic Involvement Scale (see Appendix B). Participants also reported EDI history over the previous five years. A history of extradyadic involvement was expected to correlate at least moderately with more permissive EDI beliefs.
Dating History (see Appendix I). Treas and Giesen (2000) used number of prior sexual partners as an indicator of EDI opportunity, reasoning that a more extensive sexual history would predict greater access to extradyadic partners. For a similar reason, an
extensive dating history might also predict greater EDI opportunity, as the attraction of both dating and EDI partners likely depends upon common factors. Therefore, a separate item asked for the number of dating partners within the last five years to evaluate possible correspondence to the EOS.
Interpersonal Attractiveness Item (see Appendix K). Because individual
characteristics other than physical attractiveness might elicit attention from alternative partners, respondents also reported on their general success in attracting others. Specifically, on a 7-point scale, participants noted their agreement with the statement, “In general, I think that other people are drawn to me.” Greater agreement was expected to predict higher scores on the EOS.
NEO Personality Inventory – Extraversion Scale (Costa & McCrae, 1985). The NEO PI-R measures five domains of personality including extraversion, which is computed on the basis of 30 items yielding a T-score ranging from 0 to 192. The internal consistency,
predictive validity, discriminant validity, and convergent validity of the NEO PI-R have been established elsewhere (Costa & McCrae, 1985). Given its constituent facets of warmth, gregariousness, assertiveness, activity, excitement-seeking, and positive emotions, extraversion was expected to correlate positively with EDI opportunity, as extroverted individuals would likely encounter and attract higher numbers of alternative partners.
negatively with the EIS, as sensation seekers would likely be disinclined to regard the interpersonal scenarios as inappropriate.
Additional EIS Validity Measures. Overlap was anticipated between the Extradyadic Inappropriateness Scale and the Extradyadic Beliefs Inventory, as common factors likely account for EDI beliefs and the perceived inappropriateness of specified interpersonal scenarios. Additionally, the EIS was expected to correlate negatively with EDI history and positively with both moral traditionalism and the single survey item.
Study 2 – Longitudinal Analysis Procedure
In the longitudinal study, participants completed the Extradyadic Beliefs Inventory, Extradyadic Inappropriateness Scale, Investment Model Scale, Extradyadic Opportunity Scale, and Extradyadic Involvement Scale on four occasions, each separated by
approximately three weeks and together spanning the course of approximately nine weeks. On each occasion, students reported their current relationship status. The first assessment included a background questionnaire inquiring about relationship characteristics and other demographic variables.
Surveys administration occurred over the web via a software package called Zoomerang Pro. Prior to assessment one, participants attended an orientation session familiarizing them with the study protocol, along with use of the web-based survey system. Email prompts reminded students when to complete follow-up assessments.
investigation of EDI. The method section for study 1 provides a full description of those taking part in the investigation.
In addition to the measures described for study 1, which included the newly created instruments and the various scales evaluating their psychometric properties, the participants completed the Investment Model Scale and a Background and Relationship Characteristics questionnaire described below.
Investment Model Scale (Rusbult et al.,, 1998; see Appendix M). The Investment Model Scale measures commitment level in addition to its three proposed bases of
dependence – satisfaction level, quality of alternatives, and investment size. The subscale for each base of dependence consists of five global items along with five more specific items, but only the global items are used to compute an overall subscale score. Items include 9-point likert-type scales (0 = Do not agree at all; 4 = Agree somewhat; 8 = Agree completely), and overall subscale scores range from 0 to 40. The commitment level subscale includes seven items, such that the overall subscale score ranges from 0 to 35. The internal consistency, structural validity, convergent validity, discriminant validity, and predictive validity of the Investment Model Scale have been demonstrated elsewhere (Rusbult et al., 1998).
Background and Relationship Characteristics (see Appendix N). The background questionnaire provided information regarding demographic factors and pertinent relationship characteristics. Importantly, the questionnaire assessed changes in relationship status
Factor Analytic Strategy
For each of the new scales, a common factor analytic strategy evaluated the
interrelationships among items. With the use of five-point likert responses, item distributions likely possessed ordered categorical rather than continuous properties, thereby precluding implementation of traditional factor techniques that utilize Pearson product moment correlations as the bases of analysis (Mislevy, 1986). As a more appropriate alternative, Mplus version 3.0 (Muthen & Muthen, 1998-2006) analyzed tetrachoric correlations among items, estimating the corresponding factor solution through the use of weighted least-squares with mean and variance adjustment. Others have demonstrated the method’s appropriateness for ordered categorical data of the type observed here (Mislevy, 1986). As an additional component of the analytic strategy, promax rotation allowed for intercorrelations among factors.
A variety of criteria informed the decision of how many factors to retain. The chi-square statistic and root mean chi-squared error of approximation served as indexes of overall model fit, and the eigenvalues, scree plot, and interpretability of a given solution provided additional information about how effectively it accounted for the observed pattern of
cross-loadings. Repetition of the process yielded an interpretable factor structure with each item exhibiting a strong loading on one defining factor and weak loadings on any remaining factors. The following sections present results for the EBI, EOS, and EIS, respectively. Factor Analysis of the Extradyadic Beliefs Inventory
Table 1 displays descriptive statistics for the original 37 items of the Extradyadic Beliefs Inventory. A preliminary EFA included all items, despite the existence of possible floor effects for statements 28 and 29. As formulated above, the factor analytic procedure ultimately supported the retention of four factors, and together they reflected
interrelationships among 19 items that manifested unambiguous factor loading patterns. Table 2 presents the items themselves along with their corresponding loadings on each of the four retained factors. As listed, the first factor seemingly reflected a number of different incentives that one might garner through EDI, with a representative item reading, “Itwould be fun to be physically intimate with someone other than my partner.” The second factor, which comprised a total of eight items, presumably represented the moral and ethical values an individual holds with regard to EDI. As an example, one item read as follows, “If I were physically intimate with someone other than my partner, I might never forgive myself.” Regarding the third factor, items appeared to reflect the different justifications that one might give for his/her own extradyadic behavior, as illustrated by the statement, “If I were
dissatisfied in my relationship, it would be acceptable for me to be physically intimate with someone else.” And lastly, a fourth factor, labeled “peers,” seemed to measure the attitudes of one’s friends toward EDI, with a representative item reading “My friends would
four sub-scales demonstrated a high level of internal consistency (Range of coefficient alpha = .79 - .85).
Table 1. Means and standard deviations for the EBI items
1. Under some conditions, I might be physically intimate with someone
other than my partner. 1.78 1.11
2. I were physically intimate with someone other than my partner, I would
not feel guilty. 1.50 0.97
3. If I were physically intimate with someone other than my partner, I
would like myself less. 3.83 1.22
4. Monogamy is not that important in a relationship. 1.39 0.70
5. If I were physically intimate with someone other than my partner, others
would disapprove of me. 3.96 0.89
6. I could enhance my current relationship by being physically intimate
with someone other than my partner. 1.34 0.65
7. If I were physically intimate with someone besides my partner, others
would dislike me. 3.47 1.00
8. Being physically intimate with someone else wouldn't make me a bad
boyfriend/girlfriend. 1.57 1.02
9. Others would forgive my for being physically intimate with someone
other than my partner. 3.15 0.99
10. I would lose some important friendships if I were physically intimate
with someone besides my partner. 3.23 1.22
11. It would be fun to be physically intimate with someone other than my
partner. 2.31 1.27
12. Being physically intimate with someone other than my partner would
cause me to think about myself in a negative way. 4.00 0.97
13. If I were physically intimate with someone other than my partner, I
would be a bad person. 3.36 1.11
14. If I were physically intimate with someone other than my partner, I
might never forgive myself. 3.27 1.26
15. If I were physically intimate with someone other than my partner, I
might feel better about myself. 1.63 0.82